October 1971
On the Concept of Optical Texture
J. J. Gibson, Cornell University
The World Wide Web distribution of James Gibson’s “Purple Perils” is for scholarly use with the understanding that Gibson did not intend them for publication. References to these essays must cite them explicitly as unpublished manuscripts. Copies may be circulated if this statement is included on each copy.
In the Visual World (1950) I distinguished between what I called “retinal” texture and “surface” texture. The latter was the microstructure of a surface, and I asserted that there was a micro- structure of the light reflected from a surface as well as the surface itself. I assumed that there was a one-to-one correspondence between the elements of the optical texture and those of the surface, never doubting that a retinal image was a projection of the outer worldña projection in the sense of that term that I did not carefully examine but simply borrowed from geometry and optics.
This notion of a projective correspondence between two kinds of texture was at least an advance on earlier notions of the retinal image. One could begin to speak of perspective transformation of an optical texture not just a “form” and could show how the density went along with increasing distance of the ground, and how the compression of this texture (“foreshortening”) went along with the increasing slant of a surface. In these respects, one could hypothesize that depth and distance were specified in the retinal image, contrary to the doctrine that since the third dimension was “lost” in the image it had to be “restored.”
Unhappily, however, this notion of the projective correspondence between two kinds of texture is not adequate to the whole problem of perception. There is not just one kind of “texture” of a surface; there is a “pigment structure” and a “layout structure” (Perceptual Systems, 1966, p. 208 ff). And there are still other characteristics of a surface that are specified in light. Spots of pigment or speckles on a flat surface do correspond projectively to something like spots or speckles in an optic array but the notion of the one-to-one correspondence cannot be pushed farther than that. When bumps and hollows on the surface become deep enough, projective correspondence fails. It is much too simple to assume that a pictorial array (or retinal image) is just a “projection” of the outer world.
Here is a list of cases where the elements of the optical texture fail to correspond with the elements of the surface in the environment.
1. Only facing surfaces are projected at all; occluded surfaces are not projected to the point of observation.
2. In the case of a transparent (but semi-textured) surface in front of another there are two surfaces “projected” to the point of observation, but in a sense of the tern quite different from above.
3. Cast shadows on a surface (as distinguished from the attached shadows arising from bumps and hollows) are independent of the texture of the surface (its layout structure and pigment structure) but yield an optical texture in the array.
4. Specular reflections of the source of light that bounce off a surface cause elements of the optical texture that do not correspond to elements of the structure of the surface (e.g. multiple reflections, or the scintillating reflections from water).
5. Luminous elements in a surface (radiating rather than reflecting light) will cause an optical texture that does not correspond to the texture of the surface.
If a point-to-point correspondence does not hold between the world and the optic array, what kind of correspondence does hold? There is bound to be a specificity of some sort, and the task of ecological optics is to find it. The best line to pursue at the present seems to follow from the assumption that there are no elements either in the world or in the array but nested units instead. The kinds of correspondence between the two would then have to be in terms of features or “formless invariants” (e.g. the invariants by which a caricature can specify a face better than a projective a photograph can). Moreover the normal optic array should be taken as an array at a moving point of the observation, not a “station point,” so that we can begin to understand not only the perception of surfaces that are at a slant but also the perception of surfaces that are occluded.